Skip to content
Software Architecture Guide

Fault Detection Tactic

Explanation

Section titled “Explanation”

Before a system can act on a fault, it must detect or anticipate that fault. Detection therefore comes first in availability work. It is one group of tactics. Recovery and prevention are the other groups, and recovery tactics follow detection (see Availability Tactic Categories).

The main detection techniques are:

Monitor. A monitor is a separate component. It watches the health of other parts of the system.

Ping/echo. One node sends an asynchronous request to another and waits for a reply. This checks that the node is reachable and shows the round-trip delay. See Ping/Echo.

Heartbeat. A system monitor and the process it watches exchange messages at regular intervals. See Heartbeat.

Time stamp. This tactic detects events that arrive in the wrong order. It is used mainly in distributed systems that pass messages.

Condition monitoring. The system checks conditions in a process or device. It can also check that assumptions made at design time still hold.

Voting. Several components process the same input and their outputs are compared. The most common form is triple modular redundancy: three components, identical inputs, compared outputs. In the circuit version, a majority gate outputs 1 when at least two of three inputs are 1. One failed circuit is outvoted. See Voting.

Exception detection. The system notices a condition that changes the normal flow of execution. Handling that condition is a separate topic (see Exception Handling Tactic).

Example

Section titled “Example”

An online shop has an order service and a payment service. A monitor sends a ping to the payment service at regular intervals. If no reply arrives in time, the monitor raises an alert. Recovery tactics can then take over, ideally before shoppers notice a problem.

A monitor detects a fault and triggers a recovery tactic.

Diagram source 
flowchart LR
  M[Monitor] -- ping --> S[Service]
  S -- echo --> M
  M -- no timely reply --> D[Fault detected]
  D --> R[Recovery or prevention tactic]

Checks

Section titled “Checks”
  • The document names which detection techniques are used (for example ping/echo, heartbeat, monitor, voting).
  • Where a monitor is used, it is a separate component that observes the health of other parts.
  • Where triple modular redundancy is used, the document states that three components get identical inputs and their outputs are compared.
  • Detection is linked to recovery tactics that act after a fault is detected.
Section titled “Related”

Sources

Section titled “Sources”